IntersectTermsEnum(CompiledAutomaton compiled, BytesRef startTerm) throws IOException {
super();
// if (TEST) System.out.println("Enum init, startTerm=" + startTerm);
this.fst = dict;
this.fstReader = fst.getBytesReader();
this.fstOutputs = dict.outputs;
this.fsa = compiled.runAutomaton;
this.level = -1;
this.stack = new Frame[16];
for (int i = 0; i < stack.length; i++) {
this.stack[i] = new Frame();
}
Frame frame;
frame = loadVirtualFrame(newFrame());
this.level++;
frame = loadFirstFrame(newFrame());
pushFrame(frame);
this.meta = null;
this.metaUpto = 1;
this.decoded = false;
this.pending = false;
if (startTerm == null) {
pending = isAccept(topFrame());
} else {
doSeekCeil(startTerm);
pending = !startTerm.equals(term) && isValid(topFrame()) && isAccept(topFrame());
}
}
/**
* @param input input tokenstream
* @param synonyms synonym map
* @param ignoreCase case-folds input for matching with {@link Character#toLowerCase(int)}. Note,
* if you set this to true, it's your responsibility to lowercase the input entries when you
* create the {@link SynonymMap}
*/
public SynonymFilter(TokenStream input, SynonymMap synonyms, boolean ignoreCase) {
super(input);
this.synonyms = synonyms;
this.ignoreCase = ignoreCase;
this.fst = synonyms.fst;
if (fst == null) {
throw new IllegalArgumentException("fst must be non-null");
}
this.fstReader = fst.getBytesReader();
// Must be 1+ so that when roll buffer is at full
// lookahead we can distinguish this full buffer from
// the empty buffer:
rollBufferSize = 1 + synonyms.maxHorizontalContext;
futureInputs = new PendingInput[rollBufferSize];
futureOutputs = new PendingOutputs[rollBufferSize];
for (int pos = 0; pos < rollBufferSize; pos++) {
futureInputs[pos] = new PendingInput();
futureOutputs[pos] = new PendingOutputs();
}
// System.out.println("FSTFilt maxH=" + synonyms.maxHorizontalContext);
scratchArc = new FST.Arc<>();
}
private Long lookupPrefix(BytesRef scratch, Arc<Long> arc) throws /*Bogus*/ IOException {
assert 0 == fst.outputs.getNoOutput().longValue();
long output = 0;
BytesReader bytesReader = fst.getBytesReader(0);
fst.getFirstArc(arc);
byte[] bytes = scratch.bytes;
int pos = scratch.offset;
int end = pos + scratch.length;
while (pos < end) {
if (fst.findTargetArc(bytes[pos++] & 0xff, arc, arc, bytesReader) == null) {
return null;
} else {
output += arc.output.longValue();
}
}
return output;
}
static <T> void walk(FST<T> fst) throws IOException {
final ArrayList<FST.Arc<T>> queue = new ArrayList<>();
final BitSet seen = new BitSet();
final FST.BytesReader reader = fst.getBytesReader();
final FST.Arc<T> startArc = fst.getFirstArc(new FST.Arc<T>());
queue.add(startArc);
while (!queue.isEmpty()) {
final FST.Arc<T> arc = queue.remove(0);
final long node = arc.target;
// System.out.println(arc);
if (FST.targetHasArcs(arc) && !seen.get((int) node)) {
seen.set((int) node);
fst.readFirstRealTargetArc(node, arc, reader);
while (true) {
queue.add(new FST.Arc<T>().copyFrom(arc));
if (arc.isLast()) {
break;
} else {
fst.readNextRealArc(arc, reader);
}
}
}
}
}
FSTTermsEnum(FST<Long> fst) {
this.fst = fst;
in = new BytesRefFSTEnum<>(fst);
bytesReader = fst.getBytesReader();
}
@Override
public SortedSetDocValues getSortedSet(FieldInfo field) throws IOException {
final FSTEntry entry = fsts.get(field.name);
if (entry.numOrds == 0) {
return DocValues.emptySortedSet(); // empty FST!
}
FST<Long> instance;
synchronized (this) {
instance = fstInstances.get(field.name);
if (instance == null) {
data.seek(entry.offset);
instance = new FST<>(data, PositiveIntOutputs.getSingleton());
if (!merging) {
ramBytesUsed.addAndGet(instance.ramBytesUsed());
fstInstances.put(field.name, instance);
}
}
}
final BinaryDocValues docToOrds = getBinary(field);
final FST<Long> fst = instance;
// per-thread resources
final BytesReader in = fst.getBytesReader();
final Arc<Long> firstArc = new Arc<>();
final Arc<Long> scratchArc = new Arc<>();
final IntsRefBuilder scratchInts = new IntsRefBuilder();
final BytesRefFSTEnum<Long> fstEnum = new BytesRefFSTEnum<>(fst);
final ByteArrayDataInput input = new ByteArrayDataInput();
return new SortedSetDocValues() {
final BytesRefBuilder term = new BytesRefBuilder();
BytesRef ordsRef;
long currentOrd;
@Override
public long nextOrd() {
if (input.eof()) {
return NO_MORE_ORDS;
} else {
currentOrd += input.readVLong();
return currentOrd;
}
}
@Override
public void setDocument(int docID) {
ordsRef = docToOrds.get(docID);
input.reset(ordsRef.bytes, ordsRef.offset, ordsRef.length);
currentOrd = 0;
}
@Override
public BytesRef lookupOrd(long ord) {
try {
in.setPosition(0);
fst.getFirstArc(firstArc);
IntsRef output = Util.getByOutput(fst, ord, in, firstArc, scratchArc, scratchInts);
term.grow(output.length);
term.clear();
return Util.toBytesRef(output, term);
} catch (IOException bogus) {
throw new RuntimeException(bogus);
}
}
@Override
public long lookupTerm(BytesRef key) {
try {
InputOutput<Long> o = fstEnum.seekCeil(key);
if (o == null) {
return -getValueCount() - 1;
} else if (o.input.equals(key)) {
return o.output.intValue();
} else {
return -o.output - 1;
}
} catch (IOException bogus) {
throw new RuntimeException(bogus);
}
}
@Override
public long getValueCount() {
return entry.numOrds;
}
@Override
public TermsEnum termsEnum() {
return new FSTTermsEnum(fst);
}
};
}
@Override
public SortedDocValues getSorted(FieldInfo field) throws IOException {
final FSTEntry entry = fsts.get(field.name);
FST<Long> instance;
synchronized (this) {
instance = fstInstances.get(field.name);
if (instance == null) {
data.seek(entry.offset);
instance = new FST<>(data, PositiveIntOutputs.getSingleton());
if (!merging) {
ramBytesUsed.addAndGet(instance.ramBytesUsed());
fstInstances.put(field.name, instance);
}
}
}
final NumericDocValues docToOrd = getNumeric(field);
final FST<Long> fst = instance;
// per-thread resources
final BytesReader in = fst.getBytesReader();
final Arc<Long> firstArc = new Arc<>();
final Arc<Long> scratchArc = new Arc<>();
final IntsRefBuilder scratchInts = new IntsRefBuilder();
final BytesRefFSTEnum<Long> fstEnum = new BytesRefFSTEnum<>(fst);
return new SortedDocValues() {
final BytesRefBuilder term = new BytesRefBuilder();
@Override
public int getOrd(int docID) {
return (int) docToOrd.get(docID);
}
@Override
public BytesRef lookupOrd(int ord) {
try {
in.setPosition(0);
fst.getFirstArc(firstArc);
IntsRef output = Util.getByOutput(fst, ord, in, firstArc, scratchArc, scratchInts);
term.grow(output.length);
term.clear();
return Util.toBytesRef(output, term);
} catch (IOException bogus) {
throw new RuntimeException(bogus);
}
}
@Override
public int lookupTerm(BytesRef key) {
try {
InputOutput<Long> o = fstEnum.seekCeil(key);
if (o == null) {
return -getValueCount() - 1;
} else if (o.input.equals(key)) {
return o.output.intValue();
} else {
return (int) -o.output - 1;
}
} catch (IOException bogus) {
throw new RuntimeException(bogus);
}
}
@Override
public int getValueCount() {
return (int) entry.numOrds;
}
@Override
public TermsEnum termsEnum() {
return new FSTTermsEnum(fst);
}
};
}